I have a violin that sends both audio and midi to the computer and am also using another microphone in the same vicinity. I'm trying to use the midi note value of the violin as a notch filter to filter out the sound of the string's acoustic vibration:

I can hear that this attenuates the sound a little bit, but not by that much. Any ideas for improving this? I tried adding additional vcf~ for overtones, but that didn't make a noticeable difference.

@s.elliot.perez what happens when you lower the q? Q=64 is a quarter of half-step wide!

Also, can you tolerate latency? I wrote an FFT "track spacer" in Pd to surgically duck a back track underneath live vocals that worked pretty well. If you sent the violin to the main channel and generated a sine wave from the MIDI for the side chain, I wonder if that would work?

I tried lowering the q to 10, but this didn(t seem to help.

I also tried setting an osc to be subtracted, but the osc sound ends up persisting.

Yeah, I'd be willing to put up with some latency- please share your method?

@s.elliot.perez
fftDucker~.pd
fftDuckerTest.pd
The test hasn't been tested (ha ha) but fftDucker~ is the same abstraction I used in performance. I just wrote the test so you could see how to hook up and use fftDucker~. Let me know if it works and/or does anything. I set all the constants for the female vocalists I was working with, but you can change them inside fftDucker~. The window size is set inside of [pd fft-analysis] and the range of frequencies that could potentially be suppressed are set in $0_freqMask. The increasing the former will increase both the frequency precision as well as the latency, and you will probably want to restrict the latter to the violin part's fundamental frequency range.

@jameslo Hey, thanks for that! It attenuates them more than the vcf~ method... but between the overtones and noise, the string sound is still recognizable... and the range of the filter is so wide that normal speech ends up sounding very thinned. I'd assume you're supposed to be able to set the width of the filter range, but anything above 0 immediately opens a big notch.

@s.elliot.perez You might have more success by noise cancellation..... a second microphone on the violin (not too close.... you need a similar sound to that which arrives at the vocal microphone) mixed out of phase with your voice microphone.
I suppose it would look something like......

If there is a way to automate the delay continuously with [bonk~] and automate the setting of the feedback level before speaking then that would be useful.

Unless playing harmonics the sound of a violin is seriously complex, and even if you manage to strip out the fundamental it is all the rest that makes up the violin sound that you don't want.
Otherwise, and maybe better (or maybe not), there are noise cancelling microphones that have an array of microphones processed so as to capture only your voice. They are very cheap, but I think always Bluetooth or usb.
David.